A Real-Time, GPU-Based Implementation of Aperture Domain Model Image REconstruction

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Computational Imaging

Background:

• Multipath and off-axis scattering degrade ultrasound image quality.
• Existing methods like ADMIRE are computationally intensive, limiting real-time application.
• Graphics Processing Units (GPUs) offer massive parallel processing for accelerating complex algorithms.

Purpose of the Study:

• To develop a GPU-based implementation of ADMIRE for real-time ultrasound imaging.
• To accelerate the ADMIRE algorithm for practical clinical use.
• To enable simultaneous display of ADMIRE with other beamforming techniques.

Main Methods:

• Developed a GPU-based implementation of the ADMIRE algorithm.
• Integrated the GPU-ADMIRE with a Verasonics Vantage 128 ultrasound research system.
• Benchmarked performance against a serial CPU implementation and evaluated real-time feasibility.

Main Results:

• Achieved a two-orders-of-magnitude speedup on a single GPU compared to CPU.
• Demonstrated real-time imaging capabilities through system integration.
• Successfully computed and displayed ADMIRE alongside Delay-and-Sum (DAS) and Short-Lag Spatial Coherence (SLSC) beamforming.

Conclusions:

• GPU acceleration makes ADMIRE feasible for real-time ultrasound imaging.
• The developed implementation significantly enhances image quality by suppressing clutter.
• Open-source code is provided, facilitating further research and application of ADMIRE.